Current video coders (MPEG, H.264, etc.) use a block-wise representation of the video sequence. The images are split up into macro-blocks, each macro-block is itself split up into blocks and each block, or macro-block, is coded by intra-image or inter-image prediction. Thus, certain images are coded by spatial prediction (intra prediction), while other images are coded by temporal prediction (inter prediction) with respect to one or more coded-decoded reference images, with the aid of a motion compensation known by the person skilled in the art.
For each block there is coded a residual block, also called prediction residual, corresponding to the original block decreased by a prediction. The residual blocks are transformed by a transform of discrete cosine transform (DCT) type, and then quantized with the aid of a quantization for example of scalar type. Coefficients, some of which are positive and others negative, are obtained on completion of the quantization step. They are thereafter traversed in an order of reading, generally zig-zag (as in the JPEG standard), thereby making it possible to utilize the significant number of zero coefficients in the high frequencies. On completion of the aforementioned traversal, a one-dimensional list of coefficients is obtained, which will be called “quantized residual”. The coefficients of this list are then coded by an entropy coding.
The entropy coding (for example of arithmetical coding or Huffman coding type) is performed in the following manner:
an item of information is coded entropically to indicate the location of the last non-zero coefficient of the list,
for each coefficient situated before the last non-zero coefficient, an item of information is coded entropically to indicate whether the coefficient is or is not zero,
for each previously indicated non-zero coefficient, an item of information is coded entropically to indicate whether the coefficient is or is not equal to one,
for each non-zero coefficient not equal to one situated before the last non-zero coefficient, an amplitude item of information (absolute value of the coefficient decreased by two) is coded entropically,
for each non-zero coefficient, the sign which is allotted to it is coded by a ‘0’ (for the + sign) or a ‘1’ (for the − sign).
According to the H.264 technique for example, when a macroblock is split up into blocks, a data signal, corresponding to each block, is transmitted to the decoder. Such a signal comprises:                the quantized residuals contained in the aforementioned list,        information representative of the mode of coding used, in particular:                    the mode of prediction (intra prediction, inter prediction, default prediction carrying out a prediction for which no item of information is transmitted to the decoder (“skip”));            information specifying the type of prediction (orientation, reference image, etc.);            the type of partitioning;            the type of transform, for example 4×4 DCT, 8×8 DCT, etc.            the motion information if necessary;            etc.                        
The decoding is done image by image, and for each image, macroblock by macroblock. For each partition of a macroblock, the corresponding elements of the stream are read. The inverse quantization and the inverse transformation of the coefficients of the blocks are performed so as to produce the decoded prediction residual. Next, the prediction of the partition is calculated and the partition is reconstructed by adding the prediction to the decoded prediction residual.
The intra or inter coding by competition, such as implemented in the H.264 standard, thus relies on various items of coding information, such as those aforementioned, being set into competition with the aim of selecting the best mode, that is to say that which will optimize the coding of the partition considered according to a predetermined performance criterion, for example the bitrate/distortion cost well known to the person skilled in the art.
The information representative of the mode of coding selected is contained in the data signal transmitted by the coder to the decoder. The decoder is thus capable of identifying the mode of coding selected at the coder, and then of applying the prediction in accordance with this mode.
In the document “Data Hiding of Motion Information in Chroma and Luma Samples for Video Compression”, J.-M. Thiesse, J. Jung and M. Antonini, International workshop on multimedia signal processing, 2011, there is presented a data hiding method implemented in the course of video compression.
More precisely, it is proposed to avoid including in the signal to be transmitted to the decoder at least one competition index such as arises from a plurality of competition indices to be transmitted. Such an index is for example the index MVComp which represents an item of information making it possible to identify the motion vector predictor used for a block predicted in Inter mode. Such an index, which can equal 0 or 1, is not inscribed directly into the coded data signal, but transported by the parity of the sum of the coefficients of the quantized residual. An association is created between the parity of the quantized residual and the index MVComp. By way of example, the even value of the quantized residual is associated with the index MVComp of value 0, while the odd value of the quantized residual is associated with the index MVComp of value 1. Two cases can occur. In a first case, if the parity of the quantized residual already corresponds to that of the index MVComp that it is desired to transmit, the quantized residual is coded in a conventional manner. In a second case, if the parity of the quantized residual is different from that of the index MVComp that it is desired to transmit, there is undertaken a modification of the quantized residual in such a way that its parity is the same as that of the index MVComp. Such a modification consists in incrementing or decrementing one or more coefficients of the quantized residual by an odd value (e.g.: +1, −1, +3, −3, +5, −5, etc.) and to retain only the modification which optimizes a predetermined criterion, in this instance the aforementioned bitrate-distortion cost.
At the decoder, the index MVComp is not read from the signal. The decoder simply makes do with determining the residual conventionally. If the value of this residual is even, the index MVComp is set to 0. If the value of this residual is odd, the index MVComp is set to 1.
In accordance with the technique which has just been presented, the coefficients which undergo the modification are not always chosen in an optimal manner, so that the modification applied gives rise to disturbances in the signal transmitted to the decoder. Such disturbances are inevitably detrimental to the effectiveness of the video compression.
Moreover, the index MVComp does not constitute the most beneficial item of information to be hidden since the probabilities that this index is equal to 0 or to 1 are not equal. Consequently, if this index is coded in a conventional manner by entropy coding, it will be represented, in the compressed file to be transmitted to the decoder, by a smaller quantity of data than one bit per index MVComp transmitted. Consequently, if the index MVComp is transmitted in the parity of the quantized residual, the quantity of data thus saved is smaller than one bit per index MVComp, whereas the parity of the residual could make it possible to transport an item of information of one bit per index.
Consequently, the reduction in the signaling cost, as well as the effectiveness of the compression, are not optimal.